EVGA: Amazon's New World killed some of our RTX 3090 cards due to bad soldering

midian182

Posts: 7,143   +63
Staff member
In context: Remember the furor surrounding Amazon’s New World beta in July? It had nothing to do with the game itself, but for the fact it was killing some EVGA RTX 3090 cards. Now, the hardware maker has revealed what caused the problems: poor soldering.

There were several reports of EVGA RTX 3090 cards, along with unconfirmed reports of other cards from different generations, manufacturers, and AMD, being borked by the recent New World beta. It seems Amazon Games, which said it had found no indication of widespread issues, had not implemented a framerate limiter in the menu system, causing the in-game frame rate to shoot from 100 fps to 800 fps instantly. It released a patch that capped frames per second on the menu screen to address the issue.

EVGA later replaced the affected cards—only 24 were received—and examined the broken models to try and discover what went wrong. In a statement to PC World, the company said theories that the cards’ cooling system was somehow to blame were incorrect. “In no way shape or form, is it related to the fan controller,” said a spokesperson.

EVGA added that noise on the i2c bus caused third-party software such as HWiNFO and GPU-Z to report that the fan controller wasn’t working, even though it was. The company has released a micro-controller update that shows the fan controller working properly on updated versions of the tools.

The actual reason for the failures, according to EVGA, was simply “poor workmanship.” Specifically, the sub-par soldering around the cards' MOSFET circuits. It added that only a handful of products from an early production run in 2020 suffer from this problem and that the affected batch represents fewer than 1% of all the graphics cards sold by the company.

It seems the RTX 3090 issues did little to dampen anticipation around New World's full release. The MMO is second on Steam's most-wishlisted games chart.

Permalink to story.

 

Irata

Posts: 1,720   +2,878
How could they fix an early production run feature without knowing of the issue back then ?

If they knew about it, they wouldn‘t have needed to try and discover what went wrong, if they didn‘t how could they have fixed the issue ?
 

wiyosaya

Posts: 6,518   +4,905
:rolleyes: :facepalm:

Speaking as a former electronics technician, cold solder joints can be a real beech.

However, that is what "Burn-in" testing is for. Put the cards in a test rig and run a stressful process on them in a temperature-controlled room that is kept at 50C or higher for several days. If they are still running at the end of that time period, only then should they be shipped.

EVGA probably implementing "cost effective" measures when now, their reputation is tarnished, IMO.
 

DukeJukem

Posts: 253   +272
800fps in the game menu? Why? You can't even run csgo at that framerate with the best hardware. Pretty dumb decision on amazons' part. If my current pc started pushing 800fps it would explode.
 
800fps in the game menu? Why? You can't even run csgo at that framerate with the best hardware. Pretty dumb decision on amazons' part. If my current pc started pushing 800fps it would explode.
Less of a "decision" and more of an oversight. As the article says, they released a patch later that limited the framerate at the menu. Furthermore, I've played games that shoot up to 1000fps in menus and loading screens and have never seen adverse affects. Your card is rendering a 2D image in those cases. There's no reason your card should "explode", as you say, unless it's faulty.
 
:rolleyes: :facepalm:

Speaking as a former electronics technician, cold solder joints can be a real beech.

However, that is what "Burn-in" testing is for. Put the cards in a test rig and run a stressful process on them in a temperature-controlled room that is kept at 50C or higher for several days. If they are still running at the end of that time period, only then should they be shipped.

EVGA probably implementing "cost effective" measures when now, their reputation is tarnished, IMO.
They replaced all of the affected cards and admitted to the fault so I'd say they handled this pretty honorably.
 

wiyosaya

Posts: 6,518   +4,905
They replaced all of the affected cards and admitted to the fault so I'd say they handled this pretty honorably.
I don't disagree, however, they would have never had to handle the situation at all had they done proper burn-in testing before the cards went out the door.

Maybe they will use this experience as an opportunity for improvement.
 

Dimitriid

Posts: 1,085   +2,062
You know this highlights two points for me overall:

1) If I have the option, I like EVGA overall not because they've got good quality control because well, these stuff keeps happening, but because so far they've been good at taking ownership of the issues where many other AIB partners just dodge the issues however,

2) If I was to spend 3090 amounts of money, I would expect that same price tag in attention to detail. It's quite frankly, ridiculous that they have soldering issues on such an expensive product and this is mostly because of how much Nvidia is probably charging for the chip itself. These are most certainly *not* high end products, just high end prices.

So even considering 1) I think for the most part 2) encourages most sensible consumers to just avoid the top-of-the-line products as much as possible: you'll be ok with 30 to 40% less performance and half the price and you'd be even better if you can stay on the sub 300 USD price range still.

We would still have quality control issues like this but the fact that Nvidia an partners like EVGA cooperate to keep pushing consumers to pay ridiculous amounts for only the high end products while intentionally delaying the release of stuff like the 3050 are really starting to feel like blatant attempts to manipulate the market into just spending more and more baseline while the quality control seems to be in decline through the years.